【YOLO改进】换遍IoU损失函数之DIoU Loss（基于MMYOLO）

DIoU损失函数

论文链接：https://arxiv.org/pdf/1911.08287

DIoU损失函数（Distance Intersection over Union Loss）是一种在目标检测任务中常用的损失函数，用于优化边界框的位置。这种损失函数是IoU损失函数的改进版，其不仅考虑了边界框之间的重叠区域，还考虑了它们中心点之间的距离，从而提供更加精确的位置优化。以下是DIoU损失函数的设计原理和计算步骤的详细介绍：

设计原理

一、IoU的局限性

• IoU（Intersection over Union）损失函数主要基于预测框和真实框之间的交并比，这个比例值越大表示预测框越接近真实框。
• 但IoU损失函数在预测框和真实框没有重叠时无法提供有效的梯度信息，这限制了模型的学习效率。

二、DIoU的引入

• DIoU损失在IoU的基础上增加了中心点距离的考量，这使得即使在两个框不重叠的情况下也能有效地进行梯度下降。
• 通过考虑框的几何中心距离，DIoU损失有助于减少边界框的尺寸误差，并加速收敛。

计算步骤

一、计算IoU

• 计算两个边界框A和B的交集面积I。
• 计算两个边界框的并集面积U。
• IoU计算公式为：

二、计算框的中心距离

• 设预测框的中心为，真实框的中心为 。
• 中心点距离 d 的计算公式为：。

三、计算归一化中心距离

• 计算包围预测框和真实框的最小闭合矩形（称为最小闭合框），并求出其对角线长度。
• 归一化中心距离为 ​，这样可以确保距离的比例适应不同大小的边界框。

四、计算DIoU

• DIoU损失函数定义为：
• 其中，​ 表示中心点距离的归一化平方，这样确保了距离项在损失函数中占有合适的权重。

使用PyTorch实现DIoU计算的源代码

import torch

def diou_loss(pred_boxes, gt_boxes):
    """
    计算 DIoU 损失。
    :param pred_boxes: 预测的边界框，形状为 (batch_size, 4)，格式为 (x1, y1, x2, y2)
    :param gt_boxes: 真实的边界框，形状为 (batch_size, 4)，格式为 (x1, y1, x2, y2)
    :return: DIoU 损失值
    """
    # 计算交集的坐标
    inter_x1 = torch.max(pred_boxes[:, 0], gt_boxes[:, 0])
    inter_y1 = torch.max(pred_boxes[:, 1], gt_boxes[:, 1])
    inter_x2 = torch.min(pred_boxes[:, 2], gt_boxes[:, 2])
    inter_y2 = torch.min(pred_boxes[:, 3], gt_boxes[:, 3])

    # 计算交集的面积
    inter_area = torch.clamp(inter_x2 - inter_x1, min=0) * torch.clamp(inter_y2 - inter_y1, min=0)

    # 计算预测框和真实框的面积
    pred_area = (pred_boxes[:, 2] - pred_boxes[:, 0]) * (pred_boxes[:, 3] - pred_boxes[:, 1])
    gt_area = (gt_boxes[:, 2] - gt_boxes[:, 0]) * (gt_boxes[:, 3] - gt_boxes[:, 1])

    # 计算并集的面积
    union_area = pred_area + gt_area - inter_area

    # 计算IoU
    iou = inter_area / union_area

    # 计算中心点的坐标
    pred_center_x = (pred_boxes[:, 0] + pred_boxes[:, 2]) / 2
    pred_center_y = (pred_boxes[:, 1] + pred_boxes[:, 3]) / 2
    gt_center_x = (gt_boxes[:, 0] + gt_boxes[:, 2]) / 2
    gt_center_y = (gt_boxes[:, 1] + gt_boxes[:, 3]) / 2

    # 计算中心点距离的平方
    center_distance = (pred_center_x - gt_center_x) ** 2 + (pred_center_y - gt_center_y) ** 2

    # 计算包络框的对角线距离的平方
    enclose_x1 = torch.min(pred_boxes[:, 0], gt_boxes[:, 0])
    enclose_y1 = torch.min(pred_boxes[:, 1], gt_boxes[:, 1])
    enclose_x2 = torch.max(pred_boxes[:, 2], gt_boxes[:, 2])
    enclose_y2 = torch.max(pred_boxes[:, 3], gt_boxes[:, 3])
    enclose_diagonal = (enclose_x2 - enclose_x1) ** 2 + (enclose_y2 - enclose_y1) ** 2

    # 计算 DIoU
    diou = iou - (center_distance / enclose_diagonal)

    # DIoU 损失
    diou_loss = 1 - diou

    return diou_loss

# 示例
pred_boxes = torch.tensor([[50, 50, 90, 100], [70, 80, 120, 150]])
gt_boxes = torch.tensor([[60, 60, 100, 120], [80, 90, 130, 160]])
loss = diou_loss(pred_boxes, gt_boxes)
print(loss)

 替换DIoU损失函数(基于MMYOLO)

由于MMYOLO中没有实现DIoU损失函数，所以需要在mmyolo/models/iou_loss.py中添加DIoU的计算和对应的iou_mode，修改完以后在终端运行

python setup.py install

再在配置文件中进行修改即可。修改例子如下：

    elif iou_mode == 'diou':
        # CIoU = IoU - ( (ρ^2(b_pred,b_gt) / c^2) + (alpha x v) )

        # calculate enclose area (c^2)
        enclose_area = enclose_w**2 + enclose_h**2 + eps

        # calculate ρ^2(b_pred,b_gt):
        # euclidean distance between b_pred(bbox2) and b_gt(bbox1)
        # center point, because bbox format is xyxy -> left-top xy and
        # right-bottom xy, so need to / 4 to get center point.
        rho2_left_item = ((bbox2_x1 + bbox2_x2) - (bbox1_x1 + bbox1_x2))**2 / 4
        rho2_right_item = ((bbox2_y1 + bbox2_y2) -
                           (bbox1_y1 + bbox1_y2))**2 / 4
        rho2 = rho2_left_item + rho2_right_item  # rho^2 (ρ^2)
        ious = ious - ((rho2 / enclose_area))

修改后的配置文件（以configs/yolov5/yolov5_s-v61_syncbn_8xb16-300e_coco.py为例）

_base_ = ['../_base_/default_runtime.py', '../_base_/det_p5_tta.py']

# ========================Frequently modified parameters======================
# -----data related-----
data_root = 'data/coco/'  # Root path of data
# Path of train annotation file
train_ann_file = 'annotations/instances_train2017.json'
train_data_prefix = 'train2017/'  # Prefix of train image path
# Path of val annotation file
val_ann_file = 'annotations/instances_val2017.json'
val_data_prefix = 'val2017/'  # Prefix of val image path

num_classes = 80  # Number of classes for classification
# Batch size of a single GPU during training
train_batch_size_per_gpu = 16
# Worker to pre-fetch data for each single GPU during training
train_num_workers = 8
# persistent_workers must be False if num_workers is 0
persistent_workers = True

# -----model related-----
# Basic size of multi-scale prior box
anchors = [
    [(10, 13), (16, 30), (33, 23)],  # P3/8
    [(30, 61), (62, 45), (59, 119)],  # P4/16
    [(116, 90), (156, 198), (373, 326)]  # P5/32
]

# -----train val related-----
# Base learning rate for optim_wrapper. Corresponding to 8xb16=128 bs
base_lr = 0.01
max_epochs = 300  # Maximum training epochs

model_test_cfg = dict(
    # The config of multi-label for multi-class prediction.
    multi_label=True,
    # The number of boxes before NMS
    nms_pre=30000,
    score_thr=0.001,  # Threshold to filter out boxes.
    nms=dict(type='nms', iou_threshold=0.65),  # NMS type and threshold
    max_per_img=300)  # Max number of detections of each image

# ========================Possible modified parameters========================
# -----data related-----
img_scale = (640, 640)  # width, height
# Dataset type, this will be used to define the dataset
dataset_type = 'YOLOv5CocoDataset'
# Batch size of a single GPU during validation
val_batch_size_per_gpu = 1
# Worker to pre-fetch data for each single GPU during validation
val_num_workers = 2

# Config of batch shapes. Only on val.
# It means not used if batch_shapes_cfg is None.
batch_shapes_cfg = dict(
    type='BatchShapePolicy',
    batch_size=val_batch_size_per_gpu,
    img_size=img_scale[0],
    # The image scale of padding should be divided by pad_size_divisor
    size_divisor=32,
    # Additional paddings for pixel scale
    extra_pad_ratio=0.5)

# -----model related-----
# The scaling factor that controls the depth of the network structure
deepen_factor = 0.33
# The scaling factor that controls the width of the network structure
widen_factor = 0.5
# Strides of multi-scale prior box
strides = [8, 16, 32]
num_det_layers = 3  # The number of model output scales
norm_cfg = dict(type='BN', momentum=0.03, eps=0.001)  # Normalization config

# -----train val related-----
affine_scale = 0.5  # YOLOv5RandomAffine scaling ratio
loss_cls_weight = 0.5
loss_bbox_weight = 0.05
loss_obj_weight = 1.0
prior_match_thr = 4.  # Priori box matching threshold
# The obj loss weights of the three output layers
obj_level_weights = [4., 1., 0.4]
lr_factor = 0.01  # Learning rate scaling factor
weight_decay = 0.0005
# Save model checkpoint and validation intervals
save_checkpoint_intervals = 10
# The maximum checkpoints to keep.
max_keep_ckpts = 3
# Single-scale training is recommended to
# be turned on, which can speed up training.
env_cfg = dict(cudnn_benchmark=True)

# ===============================Unmodified in most cases====================
model = dict(
    type='YOLODetector',
    data_preprocessor=dict(
        type='mmdet.DetDataPreprocessor',
        mean=[0., 0., 0.],
        std=[255., 255., 255.],
        bgr_to_rgb=True),
    backbone=dict(
        ##使用YOLOv8的主干网络

        type='YOLOv8CSPDarknet',
        deepen_factor=deepen_factor,
        widen_factor=widen_factor,
        norm_cfg=norm_cfg,
        act_cfg=dict(type='SiLU', inplace=True)

    ),
    neck=dict(
        type='YOLOv5PAFPN',
        deepen_factor=deepen_factor,
        widen_factor=widen_factor,
        in_channels=[256, 512, 1024],
        out_channels=[256, 512, 1024],
        num_csp_blocks=3,
        norm_cfg=norm_cfg,
        act_cfg=dict(type='SiLU', inplace=True)),
    bbox_head=dict(
        type='YOLOv5Head',
        head_module=dict(
            type='YOLOv5HeadModule',
            num_classes=num_classes,
            in_channels=[256, 512, 1024],
            widen_factor=widen_factor,
            featmap_strides=strides,
            num_base_priors=3),
        prior_generator=dict(
            type='mmdet.YOLOAnchorGenerator',
            base_sizes=anchors,
            strides=strides),
        # scaled based on number of detection layers
        loss_cls=dict(
            type='mmdet.CrossEntropyLoss',
            use_sigmoid=True,
            reduction='mean',
            loss_weight=loss_cls_weight *
            (num_classes / 80 * 3 / num_det_layers)),
        # 修改此处实现IoU损失函数的替换
        loss_bbox=dict(
            type='IoULoss',
            iou_mode='diou',
            bbox_format='xywh',
            eps=1e-7,
            reduction='mean',
            loss_weight=loss_bbox_weight * (3 / num_det_layers),
            return_iou=True),
        loss_obj=dict(
            type='mmdet.CrossEntropyLoss',
            use_sigmoid=True,
            reduction='mean',
            loss_weight=loss_obj_weight *
            ((img_scale[0] / 640)**2 * 3 / num_det_layers)),
        prior_match_thr=prior_match_thr,
        obj_level_weights=obj_level_weights),
    test_cfg=model_test_cfg)

albu_train_transforms = [
    dict(type='Blur', p=0.01),
    dict(type='MedianBlur', p=0.01),
    dict(type='ToGray', p=0.01),
    dict(type='CLAHE', p=0.01)
]

pre_transform = [
    dict(type='LoadImageFromFile', file_client_args=_base_.file_client_args),
    dict(type='LoadAnnotations', with_bbox=True)
]

train_pipeline = [
    *pre_transform,
    dict(
        type='Mosaic',
        img_scale=img_scale,
        pad_val=114.0,
        pre_transform=pre_transform),
    dict(
        type='YOLOv5RandomAffine',
        max_rotate_degree=0.0,
        max_shear_degree=0.0,
        scaling_ratio_range=(1 - affine_scale, 1 + affine_scale),
        # img_scale is (width, height)
        border=(-img_scale[0] // 2, -img_scale[1] // 2),
        border_val=(114, 114, 114)),
    dict(
        type='mmdet.Albu',
        transforms=albu_train_transforms,
        bbox_params=dict(
            type='BboxParams',
            format='pascal_voc',
            label_fields=['gt_bboxes_labels', 'gt_ignore_flags']),
        keymap={
            'img': 'image',
            'gt_bboxes': 'bboxes'
        }),
    dict(type='YOLOv5HSVRandomAug'),
    dict(type='mmdet.RandomFlip', prob=0.5),
    dict(
        type='mmdet.PackDetInputs',
        meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
                   'flip_direction'))
]

train_dataloader = dict(
    batch_size=train_batch_size_per_gpu,
    num_workers=train_num_workers,
    persistent_workers=persistent_workers,
    pin_memory=True,
    sampler=dict(type='DefaultSampler', shuffle=True),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        ann_file=train_ann_file,
        data_prefix=dict(img=train_data_prefix),
        filter_cfg=dict(filter_empty_gt=False, min_size=32),
        pipeline=train_pipeline))

test_pipeline = [
    dict(type='LoadImageFromFile', file_client_args=_base_.file_client_args),
    dict(type='YOLOv5KeepRatioResize', scale=img_scale),
    dict(
        type='LetterResize',
        scale=img_scale,
        allow_scale_up=False,
        pad_val=dict(img=114)),
    dict(type='LoadAnnotations', with_bbox=True, _scope_='mmdet'),
    dict(
        type='mmdet.PackDetInputs',
        meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
                   'scale_factor', 'pad_param'))
]

val_dataloader = dict(
    batch_size=val_batch_size_per_gpu,
    num_workers=val_num_workers,
    persistent_workers=persistent_workers,
    pin_memory=True,
    drop_last=False,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        test_mode=True,
        data_prefix=dict(img=val_data_prefix),
        ann_file=val_ann_file,
        pipeline=test_pipeline,
        batch_shapes_cfg=batch_shapes_cfg))

test_dataloader = val_dataloader

param_scheduler = None
optim_wrapper = dict(
    type='OptimWrapper',
    optimizer=dict(
        type='SGD',
        lr=base_lr,
        momentum=0.937,
        weight_decay=weight_decay,
        nesterov=True,
        batch_size_per_gpu=train_batch_size_per_gpu),
    constructor='YOLOv5OptimizerConstructor')

default_hooks = dict(
    param_scheduler=dict(
        type='YOLOv5ParamSchedulerHook',
        scheduler_type='linear',
        lr_factor=lr_factor,
        max_epochs=max_epochs),
    checkpoint=dict(
        type='CheckpointHook',
        interval=save_checkpoint_intervals,
        save_best='auto',
        max_keep_ckpts=max_keep_ckpts))

custom_hooks = [
    dict(
        type='EMAHook',
        ema_type='ExpMomentumEMA',
        momentum=0.0001,
        update_buffers=True,
        strict_load=False,
        priority=49)
]

val_evaluator = dict(
    type='mmdet.CocoMetric',
    proposal_nums=(100, 1, 10),
    ann_file=data_root + val_ann_file,
    metric='bbox')
test_evaluator = val_evaluator

train_cfg = dict(
    type='EpochBasedTrainLoop',
    max_epochs=max_epochs,
    val_interval=save_checkpoint_intervals)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')